Transposon Sequencing of Brucella abortus Uncovers Essential Genes for Growth In Vitro and Inside Macrophages

Brucella abortus is a class III zoonotic bacterial pathogen able to survive and replicate inside host cells, including macrophages. Here we report a multidimensional transposon sequencing analysis to identify genes essential for Brucella abortus growth in rich medium and replication in RAW 264.7 macrophages.

ABSTRACT

Brucella abortus is a class III zoonotic bacterial pathogen able to survive and replicate inside host cells, including macrophages. Here we report a multidimensional transposon sequencing analysis to identify genes essential for Brucella abortus growth in rich medium and replication in RAW 264.7 macrophages. The construction of a dense transposon mutant library and mapping of 929,769 unique mini-Tn5 insertion sites in the genome allowed identification of 491 essential coding sequences and essential segments in the B. abortus genome. Chromosome II carries a lower proportion (5%) of essential genes than chromosome I (19%), supporting the hypothesis of a recent acquisition of a megaplasmid as the origin of chromosome II. Temporally resolved transposon sequencing analysis as a function of macrophage infection stages identified 79 genes with a specific attenuation phenotype in macrophages, at either 2, 5, or 24 h postinfection, and 86 genes for which the attenuated mutant phenotype correlated with a growth defect on plates. We identified 48 genes required for intracellular growth, including the virB operon, encoding the type IV secretion system, which supports the validity of the screen. The remaining genes encode amino acid and pyrimidine biosynthesis, electron transfer systems, transcriptional regulators, and transporters. In particular, we report the need of an intact pyrimidine nucleotide biosynthesis pathway in order for B. abortus to proliferate inside RAW 264.7 macrophages.

INTRODUCTION

Brucella abortus is a class III bacterial pathogen from the Brucella genus known for being the causative agent of brucellosis, a worldwide anthropozoonosis generating major economic losses and public health issues (1). These bacteria are Gram negative and belong to the Rhizobiales order within the class Alphaproteobacteria and share common characteristics such as the DivK-CtrA regulation network which governs cell cycle regulation (2) and unipolar growth, as observed in Agrobacterium tumefaciens and Sinorhizobium meliloti (3). A specific feature of B. abortus in comparison to other Alphaproteobacteria is its multipartite genome, composed of two replicons of 2.1 and 1.2 Mb named chromosome 1 (chr I) and chromosome 2 (chr II), respectively (4). The chr I replication origin is similar to the one of Caulobacter crescentus, while the chr II replication origin resembles those found in megaplasmids of the Rhizobiales (5, 6).

One main aspect of B. abortus infections is the ability of the bacteria to invade, survive, and proliferate within host cells, including macrophages (7). Recently, the cellular infection process of B. abortus in both RAW 264.7 macrophages and HeLa cells has been extensively characterized at the single-cell level in terms of growth and genome replication, highlighting a typical biphasic infection profile (5). Indeed, during the course of cellular infections, Brucella first enters host cells through the endosomal pathway, where it remains in a “Brucella-containing vacuole” (BCV) for several hours without proliferating while preventing the maturation of its compartment into a phagolysosome (7). During that period, typical markers such as LAMP1 are acquired (7). In most cell types, surviving bacteria are able to control the biogenesis of their vacuole into an endoplasmic reticulum (ER)-derived compartment, where they actively proliferate (7–9). This process is dependent on the type IV secretion system called VirB (10, 11). The chemical composition of the replicative BCV is difficult to study directly, but mutant Brucella strains may be used as probes to gain a better knowledge of the bacterial environment in these compartments. Screening of collections of transposon mutants for attenuated strains has generated hypotheses, such as the availability of histidine that was proposed to be limited (12), which is consistent with the ability of histidinol dehydrogenase inhibitors to impair growth of Brucella in macrophages (13). However, a major drawback of previous screenings for attenuated mutants was the size of the library, typically limited to a few thousand mutants, which does not allow saturation at the genome-wide level and therefore cannot yield quantitative results. Moreover, several interesting hypotheses regarding biosynthetic pathways required for intracellular proliferation have never been investigated.

In the present study, we have performed transposon sequencing (Tn-seq) on B. abortus both before and after the infection of RAW 264.7 macrophages by using a highly saturating transposon mutant library. This library was generated by plating B. abortus on a rich medium and was subsequently used to infect RAW 264.7 macrophages for 2, 5, and 24 h. At each stage, the transposon insertion sites were mapped to identify genes in which the transposition insertion frequency is low, suggesting that these genes are required for growth and/or survival. This approach allowed identification of genes involved in several essential processes for growth on rich medium. The temporally resolved transposon sequencing analysis allowed the identification of mutants attenuated at three postinfection (p.i.) time points. Complete and near-complete pathways required for trafficking to the ER and intracellular growth in host cells have been identified, including the ability to synthesize pyrimidines when B. abortus is growing in RAW 264.7 macrophages.

RESULTS

Identification of essential genes in B. abortus 2308.

To gain genome-wide insights into the composition of essential genes necessary for growth of B. abortus on rich medium, we carried out a Tn-seq analysis. A B. abortus 2308 library of 3 × 10⁶ random mutants was constructed using a Kan^r derivative of mini-Tn5, a mini-transposon that was previously used with Brucella (12, 14). A Pxyl promoter was present in the mini-Tn5 derivative to limit the potential polar effects (see Materials and Methods). The mini-Tn5 derivatives transpose using a conservative mechanism, and a single insertion is found in each mutant (14). Directly after mating, the library was grown on rich medium and transposon insertion sites were identified by deep sequencing (Fig. 1). We identified 929,769 insertion sites from 154,630,306 mapped reads, saturating the B. abortus genome with an insertion site every 3.5 bp, on average. To allow a genome-wide analysis independent of gene annotations, we created a simple parameter assessing the transposon insertion frequency termed R200 (see Materials and Methods; see also Fig. S1 in the supplemental material), equal to the log₁₀(number of transposon insertions + 1) found within a 200-bp sliding window (Fig. 2). According to the frequency distribution of R200 values (Fig. S2), a main peak of frequency is centered on an average value of 4.05 with a standard deviation of 0.204. Since average values are strongly influenced by extreme data and because the proportion of essential genes is very different between chr I and chr II (see below), the average R200 values for chr I and II are 3.3 and 3.8, respectively. The theoretical distribution of R200 values (Fig. S2) allows the definition of cutoffs to consider growth alteration, expressed as the number of standard deviations from the average of the main peak of R200 values. Simple statistics can also be applied to compare the R200 values of different genomic regions, as indicated in Fig. 2. As shown in Fig. 2 and the externally hosted supplemental data (SD) files (SD1 to SD14) (see Table 3) showing the transposition tolerance maps (TTMs), the essential genes and genes generating a fitness defect when mutated are very often located in the coding sequences, validating the use of R200 values. One TTM was generated for each chromosome, generating SD1 for chr I (https://figshare.com/s/bd0d4fa73ad8cf7737fe) and SD2 for chr II (https://figshare.com/s/3219dfa7ac60d1cda34f). Moreover, R200 and dense coverage offer an analysis with high resolution to identify new essential genes and domains (see below).

FIG 1.

Summary of the Tn-seq approach. A transposon mutant library was initially created in B. abortus on plates using a mini-Tn5 derivative. Three million mutants were then pooled, and the resulting suspension was split in two. The first part of the pool underwent direct sequencing of Tn5-gDNA junctions, allowing the identification of mini-Tn5 insertion sites by mapping on the genomic sequence (see Fig. S1 in the supplemental material for a detailed description of the mapping and the subsequent computing), resulting in the control data set. The second part of the pool was used to infect RAW 264.7 macrophages in three separate infections. At given time points p.i. (2 h, 5 h, and 24 h), bacteria were extracted and grown on plates, colonies were collected, and their gDNA was subsequently extracted to be sequenced as for the control, resulting in 2-, 5-, and 24-h-p.i.-specific data sets. Transposon tolerance maps (TTMs) in the form of R200 values and all postinfection lists were separately compared to the control list using the Delta-R200 method (see Materials and Methods). The number of mapped read (in millions) for each data set is displayed besides its respective mapping icon, and the numbers of insertion sites are 929 × 10³ for the control condition and 742 × 10³, 713 × 10³, and 579 × 10³ for the 2-, 5-, and 24-h-p.i. data sets, respectively.

FIG 2.

Sliding R200 values along a gene map. Statistical analysis of the R200 values is indicated in Fig. S2 in the supplemental material. This analysis indicates a main peak of R200 centered on a value of 4.05. Values of 2 (−2S), 4 (−4S), and 6 (−6S) standard deviations below 4.05 are indicated on the y axis. The region shown here is an example in which nonessential genes such as the nikBCDE operon (red) are found close to the acaD gene (BAB2_0442; green) and the fadAJ operon (yellow), contributing to fitness. In the same region, an essential gene (lysK) is also identified. The R200 values of nikBCDE are statistically different from those of acaD (P < 10⁻³) and fadAJ (P < 10⁻¹⁹), according to a Scheffé pairwise comparison test with independent samples (9 and 16 nonoverlapping windows for the nikBCDE-to-acaD and nikBCDE-to-fadAJ comparisons, respectively). The light gray horizontal line is the average R200 for chr II (3.8).

TABLE 3.

Externally hosted supplemental data

Supplemental data no.	Name	Description	Link
SD1	TTM_ctrl_chrI.txt	Transposon tolerance map (R200) for chromosome I when grown on rich medium	https://figshare.com/s/bd0d4fa73ad8cf7737fe
SD2	TTM_ctrl_chrII.txt	Transposon tolerance map (R200) for chromosome II when grow on rich medium	https://figshare.com/s/3219dfa7ac60d1cda34f
SD3	Delta-R200_2hPI_chrI.txt	Attenuation profile at 2 h p.i. for chromosome I	https://figshare.com/s/77195e0a2cc1a1933b7f
SD4	Delta-R200_2hPI_chrII.txt	Attenuation profile at 2 h p.i. for chromosome II	https://figshare.com/s/2475d4b00e6a58226e40
SD5	Delta-R200_5hPI_chrI.txt	Attenuation profile at 5 h p.i. for chromosome I	https://figshare.com/s/93ce83ccea559a0de2f7
SD6	Delta-R200_5hPI_chrII.txt	Attenuation profile at 5 h p.i. for chromosome II	https://figshare.com/s/6f604b5eba5934c392f1
SD7	Delta-R200_24hPI_chrI.txt	Attenuation profile at 24 h p.i. for chromosome I	https://figshare.com/s/ea7871157f284d31eaed
SD8	Delta-R200_24hPI_chrII.txt	Attenuation profile at 24 h p.i. for chromosome II	https://figshare.com/s/958e728387a31b8ce139
SD9	TTM_replated_ctrl_chrI.txt	Transposon tolerance map (R200) for chromosome I when grown on rich medium (prior to replating)	https://figshare.com/s/519aecf6ea1bea563510
SD10	TTM_replated_ctrl_chrII.txt	Transposon tolerance map (R200) for chromosome II when grown on rich medium (prior to replating)	https://figshare.com/s/266012d35d5780c5a1b3
SD11	TTM_replated_chrI.txt	Transposon tolerance map (R200) for chromosome I when replated on rich medium	https://figshare.com/s/9282c05218ec976c4286
SD12	TTM_replated_chrII.txt	Transposon tolerance map (R200) for chromosome II when replated on rich medium	https://figshare.com/s/c5318e37bf294ff0cdde
SD13	DD-R200_24-5hPI_chrI.txt	Attenuation at 24 h p.i. compared to 5 h p.i. for chromosome I	https://figshare.com/s/44d2126e6c24ca1214d1
SD14	DD-R200_24-5hPI_chrII.txt	Attenuation at 24 h p.i. compared to 5 h p.i. for chromosome II	https://figshare.com/s/c6a0d65386f6edb2202b
SD15	ChrI.gb	Annotated chromosome I of B. abortus 2308	https://figshare.com/s/ae37affea7e62601b553
SD16	ChrII.gb	Annotated chromosome II of B. abortus 2308	https://figshare.com/s/ddee5b11fe553d1a2052

We considered essential all genes where at least one R200 value was equal to 0 under the control condition (growth on rich medium), since the probability of such events to happen randomly was estimated to be approximately 4.10⁻¹⁵ (see Materials and Methods) (15). In order to test the validity of this analysis, we checked that genes required for supposedly essential processes were indeed scored as essential if they do not have functional paralogs. As expected, genes coding for all four RNA polymerase core subunits (α, β, β′, and ω), the housekeeping σ⁷⁰, and 51 out of the 54 ribosomal proteins were found to be essential. Additionally, the previously established essentiality of pdhS, ccrM, omp2b, divK, cckA, and chpT genes (16–20) was also confirmed.

Of the 3,419 predicted genes annotated on the B. abortus genome, 491 genes were found to be essential for in vitro culture, i.e., 14.4% of the predicted genes. This percentage is in agreement with those previously reported for other Alphaproteobacteria such as C. crescentus (12.4%), Brevundimonas subvibrioides (13.4%), and Agrobacterium tumefaciens (6.9%) (15, 21). A list of all essential genes for in vitro culture is available in Table S1 in the supplemental material.

Based on the presence of a plasmid-like replication and segregation system on chr II and differences in gene content, it has been postulated that chr II might originate from an ancestrally acquired megaplasmid (5, 22). We thus tested the distribution of essential genes between the two chromosomes of B. abortus. Accordingly, 429 out of the 2,236 genes (19%) of chr I were essential. This is 3.7 times more than the 5% found on chr II, with 62 essential genes out of 1,183. This result further supports the megaplasmid hypothesis. One can thus hypothesize that essential genes have started to be transferred from chr I to chr II but that the frequency of this transfer is not sufficient to equilibrate the proportion of essential genes on both chromosomes yet. Besides the repABC operon, essential for replication initiation and segregation (5, 6), many essential genes of chr II could have been gained by recombination events with chr I. In agreement with this hypothesis, a fraction of the essential genes of chr II are clustered, such as the BAB2_0983 to BAB2_1013 region which contains 10 essential genes potentially involved in housekeeping functions like diaminopimelate biosynthesis (dapD and dapE), cell division (fzlA), and lipopolysaccharide (LPS) export (msbA).

The high resolution of the mapping (200 bp) due to the high number of reads aligned to many unique sites allows the identification of previously unannotated essential coding sequences. Indeed, since R200 values clearly map to the position of coding sequences in many instances in the genome, a drop in R200 values in a region where no gene has been predicted could indicate that a gene is indeed present and functionally relevant. This is conceptually validated by two examples shown in Fig. 3A. Interestingly, one of the two newly identified genes codes for the antitoxin component of a homologue of the SocAB system first identified in C. crescentus (Fig. 3A) (23). In C. crescentus, SocA is a proteolytic adaptor for the degradation of the SocB toxin by the ClpXP machinery (23). The essentiality of ClpXP in C. crescentus is due to the presence of the SocAB system (23). Therefore, the essentiality of clpXP genes in B. abortus (Fig. S3) could also be due to the presence of a SocAB homologue in B. abortus. Moreover, this method also allows for the reannotation of genes as exemplified by ftsK, where the open reading frame extends beyond the current ftsK locus in 5′ and matches an essentiality region (Fig. 3B). Furthermore, thanks to the high coverage of this Tn-seq, our analysis also permits the mapping of essentiality regions in genes corresponding to protein domains, as displayed by genes showing essentiality on a fraction of their coding sequence (Fig. 3C). Taken together, these observations show that high-resolution Tn-seq could support genomic reannotations and identification of essential protein domains.

FIG 3.

Genomic reannotations and identification of essential domains according to Tn-seq. The red line represents R200 values across the genome, the thin gray line represents the mean R200 per chromosome, and the black size marker corresponds to 0.5 kb. (A) Tn-seq has allowed the identification of two previously unannotated essential genes, as exemplified by the ssrA gene (encoding transfer-messenger RNA [tmRNA], allowing proteolysis of incomplete proteins) (67) between BAB1_1419 and BAB1_1420. A coding sequence located between BAB2_0403 and BAB2_0404, carrying a DUF4065 domain and well conserved (>90% identities at the protein level) within Rhizobiales, was also found to be essential. This gene encodes the antitoxin component of a toxin/antitoxin system called SocAB in C. crescentus (see the text), and BAB2_0403 is homologous to socB. Proposed reannotations are shown with a hatched line. Old and new proposed annotations are shown. (B) Open reading frame extension and matching essentiality region strongly suggest that ftsK (BAB1_1895) was misannotated. It should be noted that this corrected annotation is supported by BLASTP of the resulting extended ftsK gene against the alphaproteobacterium model C. crescentus. Examples in panels A and B are also supported by correct annotation in other genomic sequences. (C) Tn-seq is able to identify domain-specific essentiality as shown for polA (BAB1_0120) and dnaJ (BAB1_2130). In dnaJ, the Hsp70 interaction site seems essential, while in polA, encoding the DNA polymerase I, the 5′-3′ exonuclease domain is proposed to be essential.

Tn-seq allows for the reconstruction of essential pathways, complexes, and systems. Here, we specifically focused on pathways relative to the B. abortus cell cycle, which will be divided into four categories, the replication of DNA, the growth of the envelope, the cell division, and the cell cycle regulation network. Regarding the replication of DNA, α, β, γ, δ, χ, and τ subunits of the DNA polymerase III as well as the helicase dnaB and the primase dnaG are essential for growth. None of the three ε subunits (BAB1_2072, BAB2_0617, and BAB2_0967) were scored essential, which is likely due to functional redundancy. Genes responsible for the initiation of DNA replication for chromosome I (dnaA) and for chromosome II (repC) and for the segregation of chromosome I (parA and parB) and chromosome II (repA and repB) origins were also essential. Interestingly, only one homolog of the structural maintenance of chromosome gene (smc, BAB1_0522) could be found in the genome and was not essential in our Tn-seq, suggesting either that a functional analog is present or that this function is not essential in B. abortus. Genes responsible for the synthesis of peptidoglycan precursors from fructose-6-phosphate and their export to the periplasm (namely, glmS, glmM, glmU, murA, murB, murC, murD, murE, murF, murG, mraY, and ftsW) were all scored essential. Out of three predicted class A penicillin-binding proteins (PBPs), only one (BAB1_0932) was found to be essential for growth on rich medium, as well as the only predicted class B PBP, the FtsI protein. The two other class A PBPs were either only needed for optimal growth on rich medium while not being strictly essential (BAB1_0607) or not required at all (BAB1_0114).

The entire pathway responsible for the synthesis of LPS lipid A from UDP-GlcNAc (namely, lpxA, lpxB, lpxC, lpxD, lpxK, lpxXL, and kdtA), as well as the pathway responsible for its export to the outer membrane (namely, msbA, lptA, lptB, lptC, lptD, lptE, lptF, and lptG), appears to be essential for growth on rich medium. Conversely, no gene known for being involved in the LPS core synthesis (24) was scored essential for culture on plates. Moreover, none of the genes responsible for the LPS O-chain synthesis were scored essential, with the exception of wbkC (see Discussion).

Genes involved in the export of outer membrane proteins (OMPs), such as bamA, bamD, and bamE, were essential for growth on rich medium (see Discussion). However, no predicted homolog of bamB or bamC could be found. Interestingly, none of the three homologs of the OMP periplasmic chaperone degP were scored essential, which is probably due to functional redundancy. It should be noted that no clear predicted homolog of the OMP periplasmic chaperones skp and surA could be identified in silico. Additionally, among the genes responsible for lipoprotein export to the outer membrane, both lgt and lspA were essential, but surprisingly, lnt was not. Lnt is the phospholipid/apolipoprotein transacylase that is N-acylating the N-terminal cysteine in the biogenesis of lipoproteins. The lnt gene is also not essential in B. subvibrioides, suggesting that the dispensability of lnt could be a shared feature among Alphaproteobacteria.

Genes coding for the divisome proteins, i.e., ftsZ, ftsA, ftsQ, ftsK, ftsW, ftsY, and ftsI, as well as those coding for the outer membrane invagination system, tolQRAB-pal, were all essential on plates. The cell division regulatory operon minCDE as well as ftsEX was not essential, and no predicted homologs could be found for ftsB, ftsN, and zipA.

Regarding the regulation of the bacterial cell cycle in Brucella, one of the key features of the Brucella cell cycle is the DivK-CtrA pathway, conserved among many Alphaproteobacteria (25). Most but not all members of the DivK-CtrA pathway were found to be essential for growth on plates in Tn-seq. Indeed, pdhS, divK, divL, cckA, chpT, ctrA, cpdR, and clpXP were found to be essential (Fig. S3). Presumably redundant genes for c-di-GMP synthesis (pdeA and pleD) and for DivK phosphorylation (pleC and divJ) were not essential (Fig. S3).

Taken together, these data demonstrate that Tn-seq enables the reconstitution of essential pathways in a single experiment. In particular, it allows the identification of a crucial homolog within a family of several potential paralogs.

Screening for genes required for macrophage infection.

Another main objective of this study was to identify genes specifically required for macrophage infection. For this purpose, three large-scale infections of RAW 264.7 macrophages were carried out in parallel using the transposon mutant library described above (Fig. 1). For each infection, a specific postinfection (p.i.) time point was selected in order to have a better understanding of the specific gene requirement at different stages of the cellular infection process. After each time point, bacteria were extracted from infected macrophages and grown on rich medium prior to transposon insertion site identification and R200 calculation, as explained above (Fig. 1).

Attenuation corresponds to a decrease of fitness specific to infection. Therefore, in the context of an attenuated mutant, one expects that the R200 values for the mutated gene would be lower after infection than under the control condition. Consequently, for analyzing p.i. data, the control R200 values were subtracted from the corresponding p.i. R200 for each p.i. data set. This resulted in three lists of Delta-R200 values, namely, “2 h p.i. R200 − control R200,” “5 h p.i. R200 − control R200,” and “24 h p.i. R200 − control R200,” corresponding to Delta-R200 2 h p.i. (SD3 [https://figshare.com/s/77195e0a2cc1a1933b7f] and SD4 [https://figshare.com/s/2475d4b00e6a58226e40]), Delta-R200 5 h p.i. (SD5 [https://figshare.com/s/93ce83ccea559a0de2f7] and SD6 [https://figshare.com/s/6f604b5eba5934c392f1]) and Delta-R200 24 h p.i. (SD7 [https://figshare.com/s/ea7871157f284d31eaed] and SD8 [https://figshare.com/s/958e728387a31b8ce139]), respectively. A total of 165 candidates have been identified using the Delta-R200 analysis Among these candidates, 75 were found at 2 h p.i., 98 were found at 5 h p.i., and 165 were found at 24 h p.i. (Tables 1 and 2).

TABLE 1.

Attenuated B. abortus mutants in RAW 264.7 infection at 2, 5, or 24 h p.i.^a

ORFb	Gene name	Change in R200 valuec			Predicted function
		2 h p.i.	5 h p.i.	24 h p.i.
Secretion
BAB2_0068	virB1	−	−	+	Type IV secretion system
BAB2_0067	virB2	−	−	+	Type IV secretion system
BAB2_0066	virB3	−	−	+	Type IV secretion system
BAB2_0065	virB4	−	−	+	Type IV secretion system
BAB2_0064	virB5	−	−	+	Type IV secretion system
BAB2_0063	virB6	−	−	+	Type IV secretion system
BAB2_0062	virB7	−	−	+	Type IV secretion system
BAB2_0061	virB8	−	−	+	Type IV secretion system
BAB2_0060	virB9	−	−	+	Type IV secretion system
BAB2_0059	virB10	−	−	+	Type IV secretion system
BAB2_0058	virB11	−	−	+	Type IV secretion system
BAB1_0045	tamA	−	−	+	Export of autotransporters (type V secretion system)
BAB1_0046	tamB	−	−	+	Export of autotransporters (type V secretion system)
Protein synthesis and degradation
BAB1_2087	hisE	−	−	+	Histidine biosynthesis
BAB1_2082	hisB	−	−	+	Histidine biosynthesis
BAB1_1988	hisC	−	−	+	Histidine biosynthesis
BAB1_0285	hisD	−	−	+	Histidine biosynthesis
BAB1_1399	ilvC	−	−	+	Isoleucine, leucine, and valine biosynthesis
BAB1_0096	ilvD	−	−	+	Isoleucine, leucine, and valine biosynthesis
BAB1_2158	lnt	+	+	+	Lipoprotein synthesis
BAB1_1437	pepP	+	+	+	Peptidase, Xaa-Pro aminopeptidase
BAB1_0162	ibpA	−	−	+	Chaperone
BAB1_2025		+	+	+	DnaJ-like chaperone
BAB1_1115	tgt	+	+	+	tRNA modification
BAB1_0477	rplI	−	+	+	Ribosomal protein L9
BAB1_0427		−	+	+	tRNA1(Val) A37 N6-methylase TrmN6
Nucleic acid synthesis and degradation
BAB2_0641	pyrB	−	−	+	Pyrimidine biosynthesis
BAB2_0640	pyrC	−	−	+	Pyrimidine biosynthesis
BAB1_0341	pyrD	−	−	+	Pyrimidine biosynthesis
BAB1_0673	pyrE	−	−	+	Pyrimidine biosynthesis
BAB1_2132	pyrF	−	−	+	Pyrimidine biosynthesis
BAB1_0688	pyrC2	−	+	+	Pyrimidine biosynthesis
BAB1_1695	purA	−	−	+	Purine biosynthesis
BAB1_1757	purE	+	+	+	Purine biosynthesis
BAB1_0861	purS	−	−	+	Purine biosynthesis
BAB1_0024	cmk	+	+	+	CDP synthesis from CMP
BAB1_0168	ydjH	−	−	+	Adenosine kinase (AK)
BAB1_0172	rph	−	−	+	RNase
BAB2_0643	yqgF	−	+	+	Endonuclease, resolvase family
BAB1_0003	recF	−	−	+	Recombination in response to DNA damage
BAB1_1206	queF	−	−	+	Dehydrogenase, involved in queuosine biosynthesis
Electron transfer and redox
BAB2_0727	cydB	−	−	+	Cytochrome bd
BAB2_0728	cydA	−	−	+	Cytochrome bd
BAB2_0729	cydC	−	−	+	ABC transporter, cytochrome bd biogenesis
BAB2_0730	cydD	−	−	+	ABC transporter, cytochrome bd biogenesis
BAB1_1435		−	+	+	Related to cytochrome c oxidase synthesis
BAB1_0051	pcuC	−	+	+	Incorporation of Cu(I) in cytochrome c oxidase
BAB1_0139	nfuA	−	−	+	Fe-S cluster biogenesis protein
Cell envelope
BAB1_0351	wadB	+	+	+	Envelope, LPS core synthesis
BAB1_1217	murI	−	−	+	Glutamate racemase
BAB1_1462	ampD-like	−	+	+	N-acetyl-anhydromuramyl-l-alanine amidase
Regulation
BAB1_2092	bvrR	+	+	+	Two-component regulator
BAB1_1665	rpoH2	+	+	+	RNA polymerase sigma factor
BAB1_1669		+	+	+	Signal transduction, HWE family histidine kinase
BAB2_0678	rirA	−	+	+	Transcriptional regulator, iron responsive
BAB1_1517	vtlR	−	−	+	LysR transcriptional regulator controlling sRNA expression
BAB2_0118	vjbR	−	−	+	Quorum-sensing transcriptional regulator
BAB2_0143	deoR1	−	−	+	Transcriptional regulator
BAB1_0160	ptsN	+	+	+	Phosphotransferase system (PTS), IIA component
BAB1_0638	glnE	−	−	+	Glutamine pool regulation
Transport
BAB1_1460	mntH	+	+	+	Manganese transport, ion transport
BAB2_0699	oppA	+	+	+	ABC transporter, substrate binding, oligopeptide transport
BAB2_0701	oppB	+	+	+	ABC transporter, permease, oligopeptide transport
BAB2_0702	oppC	+	+	+	ABC transporter, permease, oligopeptide transport
BAB2_0703	oppD	+	+	+	ABC transporter, ATPase
BAB1_2145	phoU	−	+	+	Phosphate transport control
BAB1_1345		−	−	+	Kef-type potassium/proton antiport protein
Metabolism
BAB2_1010	glk	−	−	+	Glucokinase
BAB1_0435	glcD	+	+	+	Glycolate oxidase
BAB1_1918	lpd	+	+	+	Dihydrolipoyl dehydrogenase
BAB1_0898	bglX	−	+	+	Beta-glucosylase-related glycosidase
BAB1_1476	pldB	−	+	+	Lysophospholipase L2
BAB1_0113	fabG	−	−	+	3-Oxoacyl-ACP reductase
BAB1_0318	gph	−	−	+	Phosphoglycolate phosphatase
Unknown functions
BAB1_1485		+	+	+	Inner membrane conserved protein (DUF475)
BAB1_1766	hfaC	+	+	+	Duplicated ATPase domains
BAB1_0478		−	−	+	Inner membrane conserved protein (DUF2232)
BAB1_1283		−	−	+	Conserved periplasmic protein (DUF192)
BAB1_2069	maf-2	−	−	+	Maf-like nucleotide binding protein

^aThe R200 values (TTMs) and the genomic maps are available as supplemental data sets (see Table 3 for a complete list).

^bThe coding sequences (ORFs) untouched by previous screenings of mutant libraries are shown in bold, including some (like those for mntH, rirA, wadB, and rpoH2) that were investigated by targeted mutagenesis (63–66).

^cFor each coding sequence (ORF), a reduced R200 value is indicated by a plus. If the corresponding mutants also displayed a lower R200 after replating on rich medium, they are reported in Table 2. If a similar R200 value was found between a given time p.i. and the control, a minus is shown.

TABLE 2.

Attenuated B. abortus mutants in RAW 264.7 infection at 2, 5, or 24 h p.i. with a growth defect on plates^a

ORF	Gene name	Change in R200 valueb			Predicted function
		2 h p.i.	5 h p.i.	24 h p.i.
Protein synthesis and degradation
BAB1_1846		−	+	+	Membrane-bound metallopeptidase
BAB1_1191	clpA	+	+	+	Protease-associated factor
BAB2_0183	hisG	−	−	+	Histidine biosynthesis, first part of the pathway
BAB2_0182	hisZ	−	−	+	Histidine biosynthesis, first part of the pathway
BAB1_1098	hisI	−	−	+	Histidine biosynthesis, first part of the pathway
BAB1_2085	hisA	+	+	+	Histidine biosynthesis, first part of the pathway
BAB1_2084	hisH	+	+	+	Histidine biosynthesis, first part of the pathway
BAB1_2086	hisF	+	+	+	Histidine biosynthesis, first part of the pathway
BAB1_0704	ksgA	−	+	+	16S rRNA methyltransferase
BAB1_1553	ychF	+	+	+	Translation-associated GTPase
BAB1_2167	truB	−	+	+	tRNA modification
BAB1_1019	rluA	−	−	+	Pseudouridylate synthase, 23S RNA-specific
BAB1_1657	dsbB	−	−	+	Disulfide bond formation in periplasm
BAB1_0962		+	+	+	Protein-l-isoAsp O-methyltransferase
Nucleic acid synthesis and degradation
BAB1_0442	purD	+	+	+	Purine biosynthesis
BAB1_0730	purN	+	+	+	Purine biosynthesis
BAB1_0857	purL	+	+	+	Purine biosynthesis
BAB1_0860	purQ	+	+	+	Purine biosynthesis
BAB1_0731	purM	+	+	+	Purine biosynthesis
BAB1_0862	purC	+	+	+	Purine biosynthesis
BAB1_0868	purB	+	+	+	Purine biosynthesis
BAB1_1824	purH	+	+	+	Purine biosynthesis
Electron transfer and redox
BAB1_0091	ccmA	+	+	+	Cytochrome c maturation
BAB1_0092	ccmB	+	+	+	Cytochrome c maturation
BAB1_0093	ccmC	+	+	+	Cytochrome c maturation
BAB1_0632	ccmE	+	+	+	Cytochrome c maturation
BAB1_0633	ccmF	+	+	+	Cytochrome c maturation
BAB1_0634	ccmH	+	+	+	Cytochrome c maturation
BAB1_0631	ccmI	+	+	+	Cytochrome c maturation
BAB2_0656	ccdA	+	+	+	Cytochrome c maturation (dsbD homolog)
BAB1_0388	ccoG	−	+	+	Cytochrome c oxidase
BAB1_0389	ccoP	−	+	+	Cytochrome c oxidase
BAB1_0392	ccoN	−	+	+	Cytochrome c oxidase
BAB1_1557		−	+	+	Cytochrome c1 family
BAB1_1559		−	+	+	Ubiquinol cytochrome c reductase, iron-sulfur component
BAB1_0739		+	+	+	Electron transport chain (ETC)-complex I subunit
BAB1_1030	gor	+	+	+	Glutathione reductase
BAB2_0476	gshA	+	+	+	Gamma-glutamylcysteine synthetase
BAB1_2135	gshB	+	+	+	Glutathione synthetase
BAB1_0855		+	+	+	Glutaredoxin-like protein
Cell envelope
BAB1_1041	mlaA	+	+	+	Predicted lipoprotein
BAB1_1040	mlaD	+	+	+	ABC transporter-associated inner membrane protein
BAB1_1038	mlaE	+	+	+	ABC transporter, permease
BAB1_1039	mlaF	+	+	+	ABC transporter, ATPase
BAB2_0076	omp10	+	+	+	Outer membrane lipoprotein
BAB1_1930	omp19	−	−	+	Outer membrane lipoprotein
BAB1_0491		+	+	+	Invasion protein B (conserved periplasmic protein)
Regulation
BAB1_0143	glnD	+	+	+	Glutamine pool regulation
BAB1_0304	cenR	+	+	+	Regulator for envelope composition
BAB1_2006	cenR	+	+	+	Transcriptional regulator (also called otpR)
BAB1_1962	gntR13	−	+	+	Transcriptional regulator
BAB1_0640	pleC	−	−	+	Histidine kinase, regulation of cell cycle
BAB1_2093	bvrS	+	+	+	Histidine kinase for BvrR regulator
BAB1_2094	hprK	+	+	+	Kinase, regulator of phosphotransferase system (PTS)
BAB1_2159	hipB	+	+	+	Transcriptional regulator
BAB1_2175	irr	+	+	+	Transcriptional regulator, ferric uptake regulator
Transport and storage
BAB1_0386	copA	+	+	+	Cation metal exporter, ATPase
BAB1_0387		+	+	+	Cation exporter-associated protein
BAB2_1079	znuA	−	−	+	ABC transporter for zinc
BAB2_1080	znuC	−	−	+	ABC transporter for zinc
BAB2_1081	znuB	−	−	+	ABC transporter for zinc
BAB2_0411		+	+	+	Conserved hypothetical
BAB2_0412	tauE	+	+	+	Sulfite exporter
BAB1_1589		−	+	+	Major facilitator superfamily (transporter)
BAB1_1045	fdx	+	+	+	Ferredoxin
Metabolism
BAB2_0366	eryI	+	+	+	Erythritol metabolism (also called rpiB)
BAB2_0367	eryH	+	+	+	Erythritol metabolism (also called tpiA2)
BAB2_0370	eryC	+	+	+	Erythritol metabolism
BAB2_1013	gpm	+	+	+	Phosphoglycerate mutase
BAB1_1761	pykM	−	−	+	Pyruvate kinase
BAB2_0513	gcvT	−	−	+	Glycine cleavage system
BAB2_0514	gcvH	−	−	+	Glycine cleavage system
BAB2_0515	gcvP	−	−	+	Glycine cleavage system
BAB1_2022	cobT	−	−	+	Cobalamin biosynthesis
BAB1_2023	cobS	−	−	+	Cobalamin biosynthesis
BAB2_1012	dapB	+	+	+	Dihydrodipicolinate reductase
BAB1_0773	gppA	+	+	+	Exopolyphosphatase
BAB1_2103		+	+	+	Nucleotidyl transferase
BAB2_0442	acaD	+	+	+	Acyl coenzyme A dehydrogenase
BAB2_0511		+	+	+	Reductase
BAB2_0668	pssA	+	+	+	Phosphatidylserine synthase
BAB1_0471		−	+	+	Oxoacyl-(acyl carrier protein) reductase
BAB2_1027	upp	−	+	+	Phosphoribosyl transferase
Unknown functions
BAB1_0084	yccA	−	−	+	Inner membrane protein
BAB1_1670		−	−	+	Hypothetical protein
BAB1_1092		−	−	+	Hypothetical protein

^aThe R200 values (TTMs) and the genomic maps are available as supplemental data sets (see Table 3 for a complete list).

^bFor each coding sequence (ORF), a reduced R200 value is indicated by a plus. If a similar R200 value was found between a given time p.i. and the control, a minus is shown. These mutants also displayed a lower R200 after replating on rich medium.

In order to validate the Delta-R200 analysis, we checked for genes known to cause an attenuation during a cellular infection when mutated. For this purpose, we have chosen one control for 2 h p.i., the response regulator bvrR, and two controls for 24 h p.i., the type IV secretion system operon virB (10, 26), required for intracellular proliferation, and vjbR, an important transcriptional activator of virB which is not part of the virB operon (27, 28). As expected, the bvrR mutants were strongly attenuated from 2 h p.i., and both virB and vjbR mutants were attenuated only at 24 h p.i. (Table 1; Fig. S4), which is in agreement with the timing required by the bacterium to reach its replicative niche and proliferate. Taken together, these data lean toward the validation of the candidates identified by the Delta-R200 analysis.

We decided to assess the validity of our data sets by testing for the reproducibility of the observed phenotypes by mutating candidate genes and testing mutant survival by the counting of CFU after infection of RAW 264.7 macrophages for 2 h. We chose four candidates, two that displayed no attenuation in Tn-seq as negative controls (omp2a and ftsK-like, corresponding to BAB1_0659 and BAB2_0709 coding sequences, respectively) and two that displayed attenuation in Tn-seq as positive controls (wadB and pgk, corresponding to BAB1_0351 and BAB1_1742, respectively). As expected, the two negative-control strains did not show any sign of attenuation in comparison to the wild-type strain, while the positive-control strains exhibited a statistically significant attenuation phenotype (Fig. 4A).

FIG 4.

Validation of Tn-seq data using reconstructed mutants. (A) CFU counting after a 2-h infection of RAW 264.7 macrophages with individual omp2a, ftsK-like, wadB, and pgk mutants. The Tn-seq data indicate that omp2a and ftsK-like mutants are fully virulent at 2 h p.i., while the wadB and pgk mutants are attenuated. The wild-type (wt) strain was used as a virulent control. *, P < 0.05; ***, P < 0.001. (B) Comparison of the Tn-seq profiles of the wadB and pgk mutants, highlighting the growth defect of the pgk mutant (below the −4S threshold shown as a hatched line, i.e., 4 standard deviations under the average of the theoretical distribution of R200 values [see Fig. S2 in the supplemental material]). The gray lines correspond to the average R200 or average Delta-R200 for the whole chromosome. (C) Images of the wild-type strain, wadB mutant, and pgk mutant colonies on rich medium, supporting the hypothesis of the pgk growth defect.

One can distinguish two categories of attenuated mutants. On the one hand, mutants can be attenuated due to a failure to perform a successful infection, but on the other hand, mutants can be attenuated due to growth impairments that are actually already observed when grown on rich medium and amplified during infection. In fact, as opposed to candidates displaying a typical attenuation profile such as the wadB mutant, others displayed attenuation in conjunction with low control R200 values, as exemplified by the pgk mutant (Fig. 4B). This second type of profile suggested growth deficiencies independent of infection. In addition, while the pgk mutant strain displayed a small but significant decrease in CFU (Fig. 4A), it was clear that the colonies were smaller in size than the wild-type colonies (Fig. 4C). Therefore, the attenuation of candidates sharing a pgk-like profile in Tn-seq is likely to be due, at least in part, to growth impairments already present under the control condition. Actually, the second round of culture taking place after the infection might simply amplify the disadvantage of clones that already display growth delays on plates.

In order to investigate the effect of replating on Tn-seq candidates, we performed a new Tn-seq experiment in which the colonies of the library of transposon mutants were collected and replated prior to sequencing instead of infecting host cells. The TTMs obtained for this control condition (SD9 [https://figshare.com/s/519aecf6ea1bea563510] and SD10 [https://figshare.com/s/266012d35d5780c5a1b3]) and the replating (SD11 [https://figshare.com/s/9282c05218ec976c4286] and SD12 [https://figshare.com/s/c5318e37bf294ff0cdde]) are available at the indicated URLs. Despite a large difference in the average R200 values between the two control experiments (3.47 for the first control and 5.21 for the second), a good correlation (r = 0.86) was found between the two data sets. By comparing the control condition with the replating, we were able to monitor the fitness loss of mutants due to a second growth on the plate, independently of any infection. Surprisingly, 54% of the candidates harboring attenuation at 2, 5, or 24 h p.i. in our initial Tn-seq also displayed a similar attenuation after replating, thus indicating that the fitness loss of those mutants could be due to a growth defect detectable after a simple replating instead of infection. Consequently, attenuated candidates displaying preexisting growth impairments (listed in Table 2) should be carefully analyzed in future investigations (see Discussion). It is striking that complete or almost complete pathways fall into this category, like the purine biosynthesis (purB, purC, purD, purF, purH, purL, purMN, and purQ) and the cytochrome c maturation (ccmABC and ccmIEFH) pathways (Table 2).

Identification of hyperinvasive mutants.

Tn-seq can theoretically highlight hyperinvasive mutants in addition to attenuated mutants. Indeed, such mutants would be expected to display higher R200 values than the control, meaning that proportionally more mutant bacteria would be found inside host cells when such genes are disrupted, thus resulting in positive Delta-R200 values. Using this criterion, only nine genes (namely wbkD, wbkF, per, gmd, wbkA, wbkE, wboA, wboB, and manB_core [BAB2_0855]) were identified, and remarkably, all of them are part of the lipopolysaccharide O-chain synthesis pathway (24). Indeed, such mutants have a rough LPS, and rough mutants are known to be more invasive than the smooth parental strain (29, 30). To confirm this using our settings, a mutant of the GDP-mannose dehydratase gene (gmd) was constructed, and CFU were counted after infection of RAW 264.7 macrophages for 2 h. As expected, the resulting strain displayed increased invasiveness, which is typical of rough strains (Fig. S5).

Identification of genes required for growth in RAW 264.7 macrophages.

The 24-h-p.i.-specific candidates mainly include genes predicted to be involved in trafficking, regulation, transport, and metabolism, including amino acid and nucleic base biosynthesis (Table 1). The biosynthesis of histidine seems to be crucial, as well as the synthesis of pyrimidines, suggesting that B. abortus cannot find or take up enough of these compounds from the ER compartments in which it is proliferating. The 24-h-p.i.-specific candidates also comprise expected virulence genes such as the virB operon (11), transcriptional regulators like vjbR (28) and vtlR (31), and the cytochrome bd biosynthesis operon cydABCD (32). When our data are compared with the list of attenuated transposon mutants from previous studies performed using various infection models and Brucella species (33–37), it is striking that only 42% of the attenuated mutants identified here had already been identified. Indeed, 33 of the 79 candidates attenuated in our time-resolved Tn-seq analysis were part of the 257 candidates compiled from the previous studies. Therefore, the Tn-seq approach reported here has generated 46 new candidates, suggesting that the comprehensive analysis of the Brucella genome might yield new insights into the genes required for a macrophage infection (see Discussion).

The pyrimidine biosynthesis pathway allows intracellular proliferation.

One of the major hits of the 24-h-p.i. data set is the pyrimidine biosynthesis (here called “pyr”) pathway. In fact, with the exception of genes already essential for culture on rich medium, all pyr biosynthesis genes became strongly attenuated at 24 h p.i. (namely, pyrB, pyrC, pyrD, pyrE, and pyrF), while none of them were affected when replated.

We further investigated the pyr pathway by creating a deletion mutant of its first nonessential gene, the ΔpyrB strain. The ΔpyrB mutant grew like the wild-type strain in rich medium (Fig. S6). We then evaluated the infectious potential of the ΔpyrB strain and its complementation strain by enumerating CFU in RAW 264.7 macrophages at both 2 h p.i. and 24 h p.i. Estimation of the intracellular growth ratio (CFU at 24 h p.i. divided by CFU at 2 h p.i.) clearly showed that the ΔpyrB strain is strongly attenuated in comparison to the wild type and the complementation strains, validating the Tn-seq profile of the mutant (Fig. 5). Consistently, deletion mutants for all other pyr genes (ΔpyrC, ΔpyrD, ΔpyrE, and ΔpyrF) behaved like the wild type when cultured in rich medium (Fig. S6) and were impaired for intracellular growth, further validating the involvement of the pyr pathway for proliferation inside RAW 264.7 macrophages (Fig. 5).

FIG 5.

Intracellular growth of pyr mutants. The pyrimidine biosynthesis pathway with the corresponding genes for each step, with aspartate (Asp) and phosphoribosylpyrophosphate (PRPP) involved in UMP synthesis. The 24 h p.i./2 h p.i. CFU ratio for each pyr mutant after infection of RAW 264.7 macrophages is shown. For each strain, the log₁₀ CFU count after a 24-h infection of RAW 264.7 macrophages was divided by the corresponding log₁₀ CFU count after a 2-h infection, resulting in a ratio depicting the evolution of the bacterial load from 2 h p.i. to 24 h p.i. Accordingly, an increased load will give a ratio of >1. Each strain was compared to the wild-type control using a Scheffé analysis (one-way analysis of variance [ANOVA]), and significant differences are indicated (***, P < 0.001; ****, P < 0.0001).

DISCUSSION

In this work, a multidimensional Tn-seq analysis was performed with B. abortus to identify genes essential for growth on rich medium as well as genes required for replication in RAW 264.7 macrophages. The analysis of mutants at different times p.i. as well as the analysis of the replated library allowed the identification of genes specifically involved in the infection model tested here, by comparison with the control condition (growth on rich medium). A similar approach could be successfully applied to many other bacterial pathogens.

As most of the current therapies against bacterial pathogens aim at targeting essential processes such as translation or cell wall biosynthesis, the collection of all essential genes for growth on rich medium generates a baseline to identify novel therapeutic targets. In this study, a total of 491 candidate genes were qualified as essential for growth on rich medium plates. Interestingly, our quantitative analysis also highlights genome sections that show a reduced fitness, and statistical comparison of any regions (inside or outside predicted genes) in the genome is thus also feasible (Fig. 2). The quantitative analysis also allows the identification of regions in which mutagenesis generates a growth defect on the control plates, and thus a category of attenuated mutants with growth defects on plates (Table 2) should be distinguished from specifically attenuated mutants (Table 1). This type of discrimination is supported by the possibility that a fraction of the attenuated mutants with growth defects on plates might be nonspecifically affected during the infection process. On the other hand, the specificity of the attenuated mutants reported in Table 1 is expected to be medium dependent and could thus be challenged by other screenings. Additionally, it would be interesting to apply Tn-seq analysis to B. abortus grown in different media. In particular, it would be informative to test chemically defined media.

While peptidoglycan biosynthesis is obviously essential, only one of the three predicted class-A penicillin-binding proteins, BAB1_0932, was found to be essential for growth on rich medium, showing that Tn-seq could allow the identification of the main functional gene among a group of paralogs. As expected, BAB1_0932 is the ortholog of Atu1341, which was also identified as essential in A. tumefaciens (21). A comparison of essential genes in B. abortus, A. tumefaciens, and B. subvibrioides also reveals interesting observations. The d-Ala-d-Ala ligase-encoding gene ddl (BAB1_1447) located near the murB-murG gene cluster involved in cell wall synthesis is not essential in B. abortus, while it is essential in A. tumefaciens and B. subvibrioides. This could be explained by the presence of a paralog for ddl (BAB1_1291, also not essential) in B. abortus. MurI, a glutamate racemase that was found to be essential in C. crescentus (15) and nonessential in A. tumefaciens (21), is actually required for B. abortus growth in RAW 264.7 macrophages but not in rich medium (see below; Table 1). Tn-seq also allowed the reshaping of essential processes in comparison to those of other bacteria, as exemplified by the bam genes, responsible for the OMP export system. Brucella possesses an incomplete OMP export system composed of only bamADE and lacking bamB and bamC, and here we showed that bamE was scored essential by Tn-seq, whereas it is not essential in Escherichia coli (38). One possibility is that bamE is functionally redundant with another gene in E. coli, while this redundancy is absent in B. abortus. At the level of the regulation network involving CtrA (see Fig. S1 in the supplemental material), it is interesting to note that cpdR, sciP, gcrA, and ccrM are identified here as essential genes, which is consistent with the previous identification of ccrM as an essential gene in B. abortus (18) but surprisingly different from that for A. tumefaciens, where they are all nonessential, except for gcrA, which is absent in A. tumefaciens C58 (21). In C. crescentus, cpdR and sciP are not essential (39, 40), while gcrA and ccrM were first reported as essential (41, 42), which was later questioned by the observation of a slow growth phenotype for C. crescentus ΔgcrA and ΔccrM strains (43). The essentiality of cpdR, sciP, gcrA, and ccrM genes might indicate that this part of the regulation network is less redundant with other cell cycle control systems in B. abortus than in C. crescentus.

Intriguingly, a few genes (mucR, sodA, pgm, and wbkC) for which a viable deletion mutant has been previously reported (44–49) are actually scored as essential in our Tn-seq analysis. A pssA gene (BAB1_0470) was also scored as essential, but the corresponding viable mutant was previously characterized (50). Interestingly, mutants for another pssA homolog (BAB2_0668) were found to be attenuated from 2 h p.i. (Table 2), suggesting that these enzymes are playing distinct roles. The absence of a mini-transposon in dispensable genes has already been observed in previous Tn-seq experiments performed on other bacterial species (51). It is also possible that in the Tn-seq protocol, suppressor mutations do not have the time to be selected, and thus these genes appear as essential only in Tn-seq. Another possibility is that these mutants have a long lag phase for growth on plates or a very low growth speed and are thus wrongly detected as essential in the Tn-seq analysis.

Remarkably, a second Tn-seq analysis assessing the effects of replating revealed that 54% of these candidates displayed fitness decreases similar to those found in infection during a second round of culture. Strikingly, 55 out of the 75 initial candidates identified at 2 h p.i. are in the second category (Table 2), since they were already affected by replating. Such observations suggest that it could be important to take into account a growth defect in culture before proposing a “specific” virulence attenuation for mutant strains. Interestingly, this does not rule out that particular mutants might have an exacerbated growth defect phenotype inside host cells. Analysis of the difference between R200 values at 24 and 5 h p.i. (SD13 [https://figshare.com/s/44d2126e6c24ca1214d1] and SD14 [https://figshare.com/s/c6a0d65386f6edb2202b]) indicates that several pur genes (purA, purB, purC, purD, purF, purH, purL, purM, purN, and purQ) have lower R200 values at 24 h p.i. than at 5 h p.i. These data suggest that even if a mutant has growth problems in a given culture medium, Tn-seq analysis at different times postinfection allows the generation of hypotheses regarding attenuation at different times postinfection. Additionally, the reference culture medium is of course important, and it would be interesting to test several rich media for growth colonies before and after infection.

When comparing our candidates to the list of 257 attenuated mutants previously available from different infection models (33–37), it was surprising that only 33 genes could be found in common (Table S3). This means that 46 additional candidates were identified by Tn-seq analysis. However, this also means that 214 attenuated mutants previously reported were not identified using Tn-seq in a RAW 264.7 macrophage infection, which is not surprising, since these screenings were done with different strains/species and different infection models. However, it should be noted that, intriguingly, out of the remaining 214 candidates, 80 were categorized here as either strictly essential for growth on rich medium or essential when replated (Table S3). It is thus possible that different strains and different culture media generate different collections of essential genes or that several attenuated mutants previously reported are actually suppressors of mutants in essential genes that display a growth defect in the infection model.

The 24-h-p.i. time point revealed several pathways involved in trafficking and metabolism. The type IV secretion system VirB was needed, but the effectors proposed to be translocated to the host cell (52) were untouched in our Tn-seq analysis, which is probably the result of functional redundancy between effectors. Indeed, under the infection conditions used here, there is usually one bacterium per infected cell, and thus trans-complementation (53) is unlikely, although it cannot be completely ruled out. Regarding metabolism, glk mutants were attenuated at 24 h p.i. (Table 2), suggesting that B. abortus might need to utilize glucose in the replication niche of RAW 264.7 macrophages, which is consistent with the requirement of glucose uptake in alternatively activated macrophages (54). The biosynthesis of histidine was strongly impacted at 24 h p.i., as previously suggested (55), and we found here that it is more specifically the second part of the pathway that is important for bacterial proliferation inside RAW 264.7 macrophages (namely, hisB, hisC, and hisD). This could be due to the fact that the first half of the histidine biosynthesis pathway (composed of hisZ, hisG, hisE, hisI, hisA, hisH, and hisF) is also responsible for the production of 5′-phosphoribosyl-4-carboxamide-5-aminoimidazole (AICAR) that contributes to purine biosynthesis. The first half of the histidine biosynthesis pathway and the purine biosynthesis pathway are both impacted when colonies are replated (Table 2). Therefore, Tn-seq suggests that genes responsible for the synthesis of histidine, at least from imidazole-glycerol-3-phosphate, are required for the proliferation of B. abortus in the endoplasmic reticulum of RAW 264.7 macrophages. The ilvC and ilvD genes, coding for enzymes involved in the biosynthesis of isoleucine and valine, are also scored as required for growth in RAW 264.7 macrophages. It is also noticeable that the glutamate racemase (MurI) is required for growth in these macrophages. This enzyme converts l-Glu to d-Glu, presumably to allow the synthesis of PG. The late attenuation of these murI mutants is consistent with a late growth of B. abortus in RAW 264.7 macrophages (5). Another major hit is the biosynthesis of pyrimidines. Indeed, Tn-seq showed that all nonessential pyr biosynthesis genes (i.e., pyrB, pyrC, pyrD, pyrE, and pyrF) were consistently attenuated at 24 h p.i. in RAW 264.7 macrophages, while none of the associated mutants displayed growth defects on rich medium. Interestingly, none of the pyr genes were impacted when replated in comparison to most genes involved in purine biosynthesis (i.e., purB, purC, purD, purH, purN, purM, and two purL homologs). This is likely due to the composition of the culture medium and the heat resistance of purines and pyrimidines during medium sterilization. The hisD, hisF, pyrB, pyrC, and pyrD genes were already hit in previous screenings for attenuated mutants (12, 56), but the pyr mutants have not been complemented and the pyrimidines biosynthesis pathway has never been investigated in B. abortus. Here we show that all the mutants in genes of the pyr pathway are attenuated for growth inside macrophages, hence validating the Tn-seq data. It should be noted that a second homolog was found for pyrC (BAB1_0688); however, a B. abortus ΔBAB1_0688 strain displays a growth defect in rich medium (Fig. S7) and attenuation at 5 h p.i. in Tn-seq (Table 1), suggesting pleiotropic defects in this mutant in comparison to the pyr mutants characterized in this work. Altogether, these results strongly suggest that the ability of B. abortus to synthesize pyrimidines in the host cell is decisive for its proliferation inside macrophages. It would be interesting to investigate the survival, trafficking, and proliferation of pyr mutants in different cell types as well as in other infection models. If the inability of the pyrB mutant to proliferate inside several intracellular niches is confirmed, this mutant might be a good candidate to start vaccinal tests.

Tn-seq data also generate unexpected observations, such as the attenuation of the lnt mutants at 2 h p.i., while lnt is not required for growth on rich medium, suggesting that a redundant function is present for growth on plates but not for short-term survival in RAW 264.7 macrophages. Alternatively, it is also possible that the activity of Lnt is dispensable in B. abortus, at least under the control condition. It is noticeable that lnt is also dispensable for growth in Francisella tularensis (57), suggesting that the dispensability of Lnt is widespread. Interestingly, our screening also revealed a role for a TamAB (BAB1_0045 and BAB1_0046) system homolog for intracellular proliferation, TamAB being proposed to be involved in the translocation of outer membrane proteins (58). These data thus open new investigation pathways to better understand the molecular processes required for B. abortus survival and growth inside host macrophages.

In conclusion, Tn-seq is a comprehensive method that allowed the identification of attenuated B. abortus mutants for macrophage infection. The high coverage of the genome with transposons has allowed the identification of essential, attenuated, and nonessential genes, as well as genes or operons required for full growth on rich medium. It would be interesting to perform such experiments on other Brucella strains as well as other host cell types (including activated macrophages and trophoblasts [59]) and using more complex infection models such as animal models, e.g., a mouse intranasal infection model (60). This would generate a fundamental knowledge of the molecular arsenal required for Brucella survival and growth in the course of infections.

MATERIALS AND METHODS

Bacterial strains and media.

The wild-type strain Brucella abortus 2308 Nal^r was cultivated in 2YT (1% yeast extract, 1.6% peptone, 0.5% NaCl). The conjugative Escherichia coli S17-1 strain was cultivated in rich medium (Luria-Bertani broth). When required, antibiotics were used at the following concentrations: ampicillin, 100 μg ml⁻¹; kanamycin, 50 μg ml⁻¹ for E. coli and 10 μg ml⁻¹ for B. abortus; nalidixic acid, 25 μg ml⁻¹.

RAW 264.7 macrophage culture.

Macrophages were cultivated in Dulbecco modified Eagle medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum (Gibco), 4.5 g liter⁻¹ glucose, 1.5 g liter⁻¹ NaHCO₃, and 4 mM glutamine at 37°C with 5% CO₂.

Mini-Tn5 Kan^r plasmid construction.

The pXMCS-2 mini-Tn5 Genta^R plasmid (61) was manipulated to exchange the gentamicin resistance cassette (Genta^R) with a Kan^r gene, using a dual joining PCR strategy. The region upstream of the Genta^R cassette was amplified by PCR from the pXMCS-2 mini-Tn5 Genta^R plasmid using primers Tn-Kan part 1 F and Tn-Kan part 1 R and fused by overlapping PCR to the Kan^r coding sequence, amplified from the pNPTS138 plasmid using primers Tn-Kan part 2 F and Tn-Kan part 2 R. This DNA fragment was then fused to the region downstream of the Genta^R cassette amplified by PCR using primers Tn-Kan part 3 F and Tn-Kan part 3 R from the pXMCS-2 mini-Tn5 Genta^R plasmid. In parallel, the pXMCS-2 mini-Tn5 Genta^R plasmid was restricted using EcoRI and NdeI to excise the Genta^R fragment. The DNA fragment bearing the new Kan^r cassette was digested with EcoRI, and NdeI was then ligated in the previously restricted pXMCS-2 mini-Tn5 Genta^R plasmid. Primers used for this construct are listed in Table S2 in the supplemental material.

Mutant library generation.

One milliliter of an overnight culture of B. abortus 2308 was mixed with 50 μl of an overnight culture of the conjugative E. coli S17-1 strain carrying the pXMCS-2 mini-Tn5 Kan^r plasmid. This plasmid possesses a hyperactive Tn5 transposase allowing the straightforward generation of a high number of transposon mutants. The resulting B. abortus transposon mutants were selected on 2YT plates (2% agar) supplemented with both kanamycin and nalidixic acid. Tn5 mutagenesis generates insertion of the transposon in only one locus per genome, as demonstrated previously for Brucella (14). Each Tn5 derivative contains a C. crescentus xyl promoter that is constitutively active in B. abortus, since when it is fused to yellow fluorescent protein (YFP) coding sequence on a pBBR1-derived plasmid, it generates a fluorescent signal of uniform intensity similar to the E. coli lac promoter fused to YFP coding sequence.

RAW 264.7 macrophage infection using the transposon mutant library.

Transposon mutants were pooled using 2YT medium, diluted in RAW 264.7 macrophage culture medium to reach a multiplicity of infection (MOI) of 50, and added to the macrophages, which were previously seeded in 6-well plates to a concentration of 1.5 × 10⁵ cells per ml. A total of 16 6-well plates were planned per time point. Macrophages were then centrifuged for 10 min at 400 × g at 4°C and subsequently incubated for 1 h at 37°C with 5% CO₂. The culture medium was then removed and replaced with fresh medium containing gentamicin at 50 μg ml⁻¹ in order to kill extracellular bacteria, and macrophages were then further incubated for 1, 4, and 23 h at 37°C with 5% CO₂. For each time postinfection (2 h, 5 h, or 24 h), culture medium was removed, each well was washed twice with phosphate-buffered saline (PBS), and macrophages were lysed using PBS–0.1% Triton X-100 for 10 min at 37°C. Macrophage lysates were then plated on 100 2YT plates per time point, each supplemented with kanamycin and incubated at 37°C for 4 days in order to obtain colonies that were collected for genomic DNA (gDNA) preparation and sequencing of Tn5-gDNA junctions.

RAW 264.7 macrophage infection and CFU counting.

The infection protocol for performing CFU counting is identical to the one described above, with the exception of the inoculum, which originates from an overnight liquid culture. After infection, infected macrophages are lysed, and the resulting extracts are cultivated on 2YT plates supplemented with kanamycin. Once grown, colonies were counted to calculate the number of colonies per ml of lysate.

Analysis of essential genes for growth on plates.

In order to assess the overall transposon insertion across the B. abortus genome, we have created a parameter called R200, defined by the log₁₀(number of Tn5 insertions + 1) for a 200-bp sliding window. This sliding window was shifted every 5 bp to generate a collection of R200 values spanning the whole genome for the control condition, i.e., bacteria on plates. Given that the B. abortus genome is 3,278,307 bp, a list of 655,662 R200 values was created, with an average value of 9,481 transposon insertions mapped per window. As previously published (15), the probability of obtaining a window of a given size with no transposon insertion event can be estimated by the following formula: P = [1 − (w/g)]ⁿ, where w is the window size, g is the genome size, and n is the number of independent Tn5 insertion events. In our case, the resulting probability was 3.8 × 10⁻¹⁵, with g = 3,278,307, w = 200, and n = 544,094. It should be noted that this value accounts only for a single window, whereas essential genes are typically characterized by a series of overlapping empty windows rather than a single 200-bp window, thus further lowering the probability of finding such profiles fortuitously. Essential genes were defined as all genes having at least one R200 value equal to 0. Defined essential genes usually have many R200 values equal to 0, as indicated in supplemental data (SD) sets 1 to 14. The TTMs can be aligned with the annotated GenBank files for chromosomes I and II (SD15 and SD16) of B. abortus 2308 using Artemis (Sanger Institute [http://www.sanger.ac.uk/science/tools/artemis]). A list of the externally hosted supplemental data sets is shown in Table 3.

Statistical analysis is described in Fig. S2. Briefly, a main frequency peak centered on an R200 value of 4.05 was used to predict a theoretical distribution of R200 values from which thresholds corresponding to 2 (−2S), 4 (−4S), and 6 (−6S) standard deviations (S) for the average of the theoretical distribution were computed.

Analysis of the effect of replating on rich medium was tested in an independent Tn-seq analysis in which a new mutant library was constructed with the same mini-Tn5 derivative as described above. All colonies were collected, and the resulting suspension was used, on the one hand, for control analysis (data in TTM_replated_ctrl_chrI and TTM_replated_ctrl_chrII) and, on the other hand, for replating on the same rich medium. Colonies generated after replating were collected and analyzed by Tn-seq as described above (data in TTM_replated_chrI and TTM_replated_chrII).

Attenuation in infection analysis.

For each postinfection sample (2 h, 5 h, and 24 h p.i.), a list of R200 values was calculated as for the control condition. Then, each R200 value from the control sample list was subtracted from each R200 value from the postinfection sample list separately, generating three Delta-R200 data sets, SD3 to SD8. Therefore, regions with a neutral Delta-R200 value have no impact during infection when mutated, while regions with a negative Delta-R200 value are attenuated during infection, and regions with a positive Delta-R200 value depict hyperinvasiveness for the corresponding mutants. For each time postinfection, the frequency distribution of Delta-R200 values was computed to define a normal distribution with an average and a standard deviation covering the main peak of this distribution. The threshold for negative Delta-R200 values was set at −0.75 log₁₀ for the “2 h p.i. R200 − control R200” and “5 h p.i. R200 − control R200” Delta-R200 analyses, selecting, respectively, 5.5% and 6.7% of windows from the total genome. The threshold for negative Delta-R200 values was set at −1 log₁₀ for the “24 h p.i. R200 − control R200” Delta-R200 analysis, allowing selection of 10.3% of the windows. The threshold for positive Delta-R200 values for the “2 h p.i. R200 − control R200” condition was set at +0.6 log₁₀, selecting 1.1% of the windows. The genes covered by selected windows were considered required for the infection, with usually most of their coding sequences covered.

Generation of the B. abortus targeted mutants.

Unless stated otherwise, all B. abortus mutants were generated by insertion of a plasmid in the targeted gene, according to a previously published procedure (62). The primer sequences used to generate PCR products cloned in the disruption plasmids are available in Table S2.

All deletion strains were constructed using a previously described allelic exchange strategy (5). The primers used to amplify upstream and downstream regions of the target genes required for homologous recombination are also available in Table S2.

Growth curves.

Growth was monitored in 2YT medium at 37°C for 72 h by measuring the optical density at 600 nm using a permanently shaking plate reader (Epoch2 microplate spectrophotometer; Biotek).